Ken Snyder (@kensnyder) header
Ken Snyder (@kensnyder) profile picture

Product GTM @Cisco | Local AI Ambassador | Award-winning nature photographer

Michigan, USA
Joined June 2008
Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)Photo by Ken Snyder (@kensnyder)
Yes plus ability to learn & optimize based on the full thinking context is a game changer.
deepseek v4.1 flash 1017 is basically opus 5.3 if it existed
48
Fantastic news!
Last week, we removed the need for a developer account and X API credits to use the X connector in Grok Bot Ask @bot to manage your bookmarks, go over your DMs, or draft an article for you free of additional charge
1
47
Stunning. Underscores the requirement for secure network architecture across the entire enterprise fabric.
OpenAI's AI agents figured out how to secretly communicate between isolated virtual machines using CPU cache timing. No internet. No network connections. No shared files. Imagine two prisoners in solitary confinement discovering they can communicate by tapping on a shared wall. Except the wall was a CPU cache.
49
Great work by the @TalosSecurity team getting this out quickly! If anything in your business runs on PHP, this is critical info and requires every company to move quickly. The old habit of baking patches into the next general purpose release isn't going to survive the AI era.
Two critical remote code execution vulnerabilities in PHP core could have implications across its wider ecosystem. Talos has published a blog to explain what we know so far, and details the Snort coverage available, including rules in the Community Ruleset blog.talosintelligence.com/t…
5
31
1
126
43,241
Looking forward to fully evaluating this one! NVIDIA just dropped a 1B multimodal embedder that might be new goto for enterprise RAG. 32K context, 2048 depth, image & text. huggingface.co/nvidia/Nemotr…
34
Only took 4 days of agents coding to run out of disk space on new Mac with 2TB drive. Resulted in 956gb of build caches across 580 worktrees while working on a single project. Disk space will be an increasing major bottleneck as agent swarms expand.
1
1
45
NVIDIA and @Cisco introduce pluggable optics to scale AI factories. USFP is the first module to support 400Gb/s per lane, delivering 4x the bandwidth density of today's optics. AI builders can connect more compute to the same switch — no extra network tiers, less power per bit moved. Learn more: nvda.ws/4B6Ch7e
67
201
85
1,758
133,126
'Scroll jank' = technical jargon gone too far.. 🤣
Replying to @ChatGPT
Delegate work in Codex to your dot. Your dot can now start work in Codex and follow up on existing threads, drawing on your ChatGPT conversations, Codex threads, and automations. It’s also better at deciding when to keep a thread going or start fresh. Your dot can also review and edit Scheduled Tasks in ChatGPT Work, too. Ask what’s scheduled or update a recurring task as plans change.
38
EmbeddingGemma 2 looks interesting for video and audio; however, how much will it be limited by only 8K context? I'm averaging 2-4k of vision extraction context for a single png screenshot from a legacy pptx, docx, pdf. Also pretty underwhelming on the text MTEB ranking. Qwen3-Embedding-0.6B last updated in July 2025 still out performs on the text MTEB?! The enterprise market needs US developed open-source top 10 mteb ranking embedding model that delivers min of 64k text context and at least 1024 dimensions while under 1GB memory to run.
Replying to @googlegemma
EmbeddingGemma 2 provides best-in-size-class performance for benchmarks like MTEB (Code) and MIEB (Lite). It’s well-suited for local codebase indexing, semantic code search, and coding agent retrieval, delivering a significant 9.92-point improvement on code performance over EmbeddingGemma.
52
Pure awe! First image should be opening scene to Interstellar sequel.
Liftoff of Starship's first orbital flight
1
55
Opus 5.5 is best Opus release since 4.6. Fable 5.1 Max does more with less context usage compared to Opus 5.5 xhigh but so far I prefer the Opus improved reflection & validation performance while working on large codebase.
1
2
74
The CEOs I talk to are asking the same three questions about AI: 1) can I trust it?; 2) can I secure it?; and 3) can I afford it? This week the @Splunk team helped answer all three. That is what critical infrastructure for the AI era looks like. Great work, and worth the watch if you missed the #splunkconf26 keynote. cs.co/6005BGpNAL
9
15
1
83
77,412
Bigger AI models aren’t always better. Discover how small language models purpose-built for security can deliver speed, efficiency, privacy, and flexibility for real-world security challenges. Join Cisco AI experts on September 22. cs.co/6012BGp63y #AISecurity
8
2
28
6,629
Same. Rust + Tauri with Codex-rs bundled in unlocks incredible simple & efficient framework for building custom personal apps
I'm seeing way bigger productivity gains from building personal desktop apps than personal web apps. My personal super desktop app is insane. Saves me hours/wk.
1
85
Ken Snyder retweeted
We are past the point where we can remain safe and secure without the use of AI. AI is a pre-requisite for safety and security because as attacks become more sophisticated and machine scale, human scale defense simply won't be able to combat the attacks due to the sheer scale differential. Security is a team sport and the only enemy is the adversary. However, the nuance that sometimes gets missed here is that unsafe acts can be performed that threaten security not just because of malicious intent but also because an agent may exercise poor judgement. So there will be a need to fuse data sets of security and observability because it will get harder and harder to distinguish a malicious attack from an act by an agent that exercises poor judgement. The other point that is important to understand is that typically lack of democratization for AI can create unsafe environments for all. Reason is that the strength of the value chain is typically dictated by the weakest link. So availability of the best defenses to all institutions affordably will be a critcal element of success. The issues we are discussing have a ton of nuance. But the elegant solutions are starting to emerge. Elon's (@elonmusk) suggestion to have models check competitive models and have a public benchmark that shows how the models perform, and have this be globally coordinated is a great suggestion. There will be many such ideas that start to emerge. While it might not feel so, having the right discussions with the appropriate level of nuance is super important at this time because two things can hold true at the same time. That security and safety risks are real, and that global competitiveness for a country is a necessary ingredient for success. What's being debated are implementation approaches which need ideation. Three things worth considering: 1. Assume AI is a pre-requisite for effective security 2. Token efficiency will be needed to ensure the right security posture. 3. Security is a team sport, and data sharing will be essential. Not sure about others, but I feel encouraged that these discussions are being had and the one area that is not debatable tis that eventually we are all on Team Humans! AND given the risks created with AI, AI will be essential for ensuring human safety.
6
5
3
59
3,426
Never forget 9/11
75
4,708
87
42,685
367,327
1923. When strategic Ford acquisition brought consistent precision measurement to the line. Exactly what agent swarms need today to deliver validated quality output.
A Ford assembly line operates in 1923, utilizing raised platforms to move chassis smoothly past workers stationed with specialized tools.
3
111
Recently used gpt to analyze my current physical office layout to optimize for efficiency and positive energy. Spectacular results. In a way, new layout functions so well that it’s effortless to keep clean. Highly recommend!
Why do i keep having the thought of asking Astra to clean my office
1
3
218
Exciting to see! Improving agent ‘reflection’ is the way: “We encouraged different groups of agents to explore a diversity of approaches. After some time, we cross-pollinated the agent groups by using Codex to consolidate the most useful insights from each agent group.”
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics. The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra. The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
54